The Elasticsearch logs will show long GC pauses if they are occurring. An alternative that gives more detail is to enable GC logging. If you can turn on even simple GC logging (-verbose:gc -XX:+PrintGCDateStamps -Xloggc:<path>) that should be more than enough to rule out or give evidence for GC.
Good to know... we can ignore it until the next release...
I may be able to enable that when we do our next rolling restart... in the mean time, I double checked... not a single mention of the term 'garbage' in our logs. Aside from the monitor error, the logs are pretty clean.
If there is anything in the logs, it can be found by the regex gc.*duration. Also, you can get a rough idea of garbage collection times by looking at _node/stats and grabbing the jvm and gc objects from the response.